| 1 |
// SPDX-License-Identifier: GPL-2.0-only |
| 2 |
/* IEEE754 floating point arithmetic |
| 3 |
* double precision: common utilities |
| 4 |
*/ |
| 5 |
/* |
| 6 |
* MIPS floating point support |
| 7 |
* Copyright (C) 1994-2000 Algorithmics Ltd. |
| 8 |
*/ |
| 9 |
|
| 10 |
#include "ieee754dp.h" |
| 11 |
|
| 12 |
union ieee754dp ieee754dp_sub(union ieee754dp x, union ieee754dp y) |
| 13 |
{ |
| 14 |
int s; |
| 15 |
|
| 16 |
COMPXDP; |
| 17 |
COMPYDP; |
| 18 |
|
| 19 |
EXPLODEXDP; |
| 20 |
EXPLODEYDP; |
| 21 |
|
| 22 |
ieee754_clearcx(); |
| 23 |
|
| 24 |
FLUSHXDP; |
| 25 |
FLUSHYDP; |
| 26 |
|
| 27 |
switch (CLPAIR(xc, yc)) { |
| 28 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN): |
| 29 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN): |
| 30 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN): |
| 31 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN): |
| 32 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN): |
| 33 |
return ieee754dp_nanxcpt(y); |
| 34 |
|
| 35 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN): |
| 36 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN): |
| 37 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO): |
| 38 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM): |
| 39 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM): |
| 40 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF): |
| 41 |
return ieee754dp_nanxcpt(x); |
| 42 |
|
| 43 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN): |
| 44 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN): |
| 45 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN): |
| 46 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN): |
| 47 |
return y; |
| 48 |
|
| 49 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN): |
| 50 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO): |
| 51 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM): |
| 52 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM): |
| 53 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF): |
| 54 |
return x; |
| 55 |
|
| 56 |
|
| 57 |
/* |
| 58 |
* Infinity handling |
| 59 |
*/ |
| 60 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF): |
| 61 |
if (xs != ys) |
| 62 |
return x; |
| 63 |
ieee754_setcx(IEEE754_INVALID_OPERATION); |
| 64 |
return ieee754dp_indef(); |
| 65 |
|
| 66 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF): |
| 67 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF): |
| 68 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF): |
| 69 |
return ieee754dp_inf(ys ^ 1); |
| 70 |
|
| 71 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO): |
| 72 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM): |
| 73 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM): |
| 74 |
return x; |
| 75 |
|
| 76 |
/* |
| 77 |
* Zero handling |
| 78 |
*/ |
| 79 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO): |
| 80 |
if (xs != ys) |
| 81 |
return x; |
| 82 |
else |
| 83 |
return ieee754dp_zero(ieee754_csr.rm == FPU_CSR_RD); |
| 84 |
|
| 85 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO): |
| 86 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO): |
| 87 |
return x; |
| 88 |
|
| 89 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM): |
| 90 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM): |
| 91 |
/* quick fix up */ |
| 92 |
DPSIGN(y) ^= 1; |
| 93 |
return y; |
| 94 |
|
| 95 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM): |
| 96 |
DPDNORMX; |
| 97 |
fallthrough; |
| 98 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM): |
| 99 |
/* normalize ym,ye */ |
| 100 |
DPDNORMY; |
| 101 |
break; |
| 102 |
|
| 103 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM): |
| 104 |
/* normalize xm,xe */ |
| 105 |
DPDNORMX; |
| 106 |
break; |
| 107 |
|
| 108 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM): |
| 109 |
break; |
| 110 |
} |
| 111 |
/* flip sign of y and handle as add */ |
| 112 |
ys ^= 1; |
| 113 |
|
| 114 |
assert(xm & DP_HIDDEN_BIT); |
| 115 |
assert(ym & DP_HIDDEN_BIT); |
| 116 |
|
| 117 |
|
| 118 |
/* provide guard,round and stick bit dpace */ |
| 119 |
xm <<= 3; |
| 120 |
ym <<= 3; |
| 121 |
|
| 122 |
if (xe > ye) { |
| 123 |
/* |
| 124 |
* Have to shift y fraction right to align |
| 125 |
*/ |
| 126 |
s = xe - ye; |
| 127 |
ym = XDPSRS(ym, s); |
| 128 |
ye += s; |
| 129 |
} else if (ye > xe) { |
| 130 |
/* |
| 131 |
* Have to shift x fraction right to align |
| 132 |
*/ |
| 133 |
s = ye - xe; |
| 134 |
xm = XDPSRS(xm, s); |
| 135 |
xe += s; |
| 136 |
} |
| 137 |
assert(xe == ye); |
| 138 |
assert(xe <= DP_EMAX); |
| 139 |
|
| 140 |
if (xs == ys) { |
| 141 |
/* generate 28 bit result of adding two 27 bit numbers |
| 142 |
*/ |
| 143 |
xm = xm + ym; |
| 144 |
|
| 145 |
if (xm >> (DP_FBITS + 1 + 3)) { /* carry out */ |
| 146 |
xm = XDPSRS1(xm); /* shift preserving sticky */ |
| 147 |
xe++; |
| 148 |
} |
| 149 |
} else { |
| 150 |
if (xm >= ym) { |
| 151 |
xm = xm - ym; |
| 152 |
} else { |
| 153 |
xm = ym - xm; |
| 154 |
xs = ys; |
| 155 |
} |
| 156 |
if (xm == 0) { |
| 157 |
if (ieee754_csr.rm == FPU_CSR_RD) |
| 158 |
return ieee754dp_zero(1); /* round negative inf. => sign = -1 */ |
| 159 |
else |
| 160 |
return ieee754dp_zero(0); /* other round modes => sign = 1 */ |
| 161 |
} |
| 162 |
|
| 163 |
/* normalize to rounding precision |
| 164 |
*/ |
| 165 |
while ((xm >> (DP_FBITS + 3)) == 0) { |
| 166 |
xm <<= 1; |
| 167 |
xe--; |
| 168 |
} |
| 169 |
} |
| 170 |
|
| 171 |
return ieee754dp_format(xs, xe, xm); |
| 172 |
} |